%----------------------------------------------------------------
% Bayesian Local Projections (BLP) SETTING
%----------------------------------------------------------------


%random walk prior
%hyperPars.isrw      =true(1,length(dataStructure.varname)); %rw prior
hyperPars.lambda    =0.4;    %tightness of VAR coeffs prior (the higher lambda the closer to OLS) lambda=0-> VAR, lambda=infinity -> OLS
hyperPars.lambdaC   =1e5;   %intercept (you want this to be large)
hyperPars.lambdaP   =0.4;    %tightness of projCoeffs prior (same role as above)
hyperPars.miu       =1;     %sum of coefficients prior (constraint multiplier)
hyperPars.theta     =2;     %cointegration prior (constraint multiplier) 
hyperPars.alpha     =2;     %lag decaying coeff for NIW prior


%set hyperpriors options (matches GLP fields); if you want default values
%(when available) set to empty [];
hyperPriorsOptions.hyperpriors   = true;                  %find optimal hyperparameters: NO default option
hyperPriorsOptions.Vc            = 1e5;                   %variance of the VAR constant (default=1e6)
%hyperPriorsOptions.pos           = find(~hyperPars.isrw); %position of stationary variables
hyperPriorsOptions.pos           = []; %position of stationary variables

hyperPriorsOptions.MNalpha       = [];                    %lag decaying coeff of NIW prior (default=2)
hyperPriorsOptions.MNpsi         = false;                 %residual variance univariate AR(1) std (default=hyperprior)
hyperPriorsOptions.noc           = false;                 %sum of coefficients prior: NO default option
hyperPriorsOptions.sur           = false;                 %cointegration prior: NO default option
hyperPriorsOptions.Fcast         = false;                 %build forecasts: NO default option
%hyperPriorsOptions.hz            = modelSpec.nHorizons;   %max forecast horizon: NO default option
hyperPriorsOptions.hz            = nH;   %max forecast horizon: NO default option
hyperPriorsOptions.mcmc          = false;                 %run metropolis-hasting algorithm: NO default option
hyperPriorsOptions.Ndraws        = 1200;                  %default=20k
hyperPriorsOptions.Ndrawsdiscard = 200;                   %default=10k
hyperPriorsOptions.MCMCconst     = 1;                     %default=1
hyperPriorsOptions.MCMCfcast     = false;                 %store forecast at each MCMC draw (default=true)
hyperPriorsOptions.MCMCstorecoeff= false;                 %store coefficients at each MCMC draw (default=true)
hyperPriorsOptions.initialValues = hyperPars;             %see above

%%% Gibb Sampling
nDraws=1200;
nBurn=200;
nJump=1;